Eleven years of waiting times in Emilia-Romagna: the 2021 step never closed
7 min read
Someone books a dermatology appointment and is told a date. That date is all they see: they do not know whether it is normal, whether a year ago it would have been closer, whether in another province it would be different. The Emilia-Romagna Region has published the indicator that answers those questions for years, one week at a time, and publishes it well. Nobody, though, keeps it in a row: the site shows the present, and the present alone cannot tell you whether something is getting worse.
We downloaded the whole archive and lined the weeks up. That is 499 weeks of data for the most representative priority class, from 6 July 2015 to 23 August 2026: eleven years. The result is not what we expected.
What the Region publishes, and what we did
The regional TdAER system publishes, for each of the eleven health authorities and for every specialist service monitored, the share of bookings served within the statutory standard and the median wait, split by priority class: B within ten days, D within thirty days for visits and sixty for diagnostics, P within one hundred and twenty.
Those numbers are the Region’s. Ours is only the recomposition: gathering every week available, aligning them on a single scale, and checking that the series holds up over time. It does: thirty-nine services are present in all eleven years, three dropped out along the way, none entered, and no code was reused under a different name. This is not three different series glued together, it is one series.
The territorial grain stops at the health authority: the finer sub-authority catchment areas do not appear in this data.
The step, and the fact that it never closed
| year | class B | class D | class P |
|---|---|---|---|
| 2016 | 99.6% | 98.0% | — |
| 2018 | 95.6% | 98.3% | — |
| 2021 | 89.7% | 88.4% | 95.0% |
| 2023 | 89.6% | 92.0% | 91.6% |
| 2025 | 89.3% | 86.8% | 87.0% |
| 2026 | 90.0% | 87.8% | 85.4% |
Between 2018 and 2021 class D loses ten points. Five years later it has not recovered them: 2025 closes lower than 2021. Class B had given way earlier, back in 2018, and has since hovered around ninety per cent without returning to earlier levels. Class P — published only from 2020, so a shorter series and not comparable with the others over the long run — is the one falling fastest now, from 95.0% in 2021 to 85.4% this year.
What makes the step interesting is something else, and it is usually missing from the debate: volume did not change. About forty-three thousand bookings a week in class D, essentially identical from 2016 to today. No wave of new demand arrived. What changed is the capacity to absorb the usual demand within the expected time.
What the percentage does not say
A share-within-standard is a binary indicator dressed up as a measure: it says how many bookings fall inside the threshold, not how far outside the excluded ones fall. That is why it has to be read next to the median wait, which the Region publishes alongside it. In the latest week available the regional median is seven days in class B, twenty in class D, thirty-five in class P. In that same week, though, the worst cell in class D sits at two hundred and fifty-six days, and in class P at six hundred and four.
Two caveats that must be stated, and that the indicator carries by construction. First: it is ex ante. It measures the wait quoted at the moment of booking, not the one the patient actually lived through — whoever gives up, moves to private care or travels to another province appears nowhere. Second: the archive has gaps, and they are the source’s gaps, not ours. Twenty-six weeks between February and August 2019, for class D alone. Twenty-nine weeks between March and October 2020, for every class: that one is the suspension of elective activity, and explains itself. The 2019 gap, six whole months, does not.
Where the proof sits today
Anyone wanting to answer the question of the person in the queue — has it always been like this? — has to assemble pieces no single system holds together. The booking system knows the date offered, not the history of that service in that area. The regional portal publishes the current week, not the series. The archive of past weeks exists, but has to be queried one week at a time. The figure for the wait actually endured, the one that matters to whoever is waiting, is not in the ex ante indicator and is not published. And the choice of those who give up — the most important of all, because it empties the lists without solving anything — is measured nowhere.
The recomposition needs no confidential data: it needs only that someone keeps the series. It is exactly the work we do on an organisation’s data, done here on data that is public and verifiable by anyone.
What we do not know
We do not know why six months of 2019 are missing. We do not know which part of the step is a supply shortfall, which is catch-up on postponed care and which is a change in how people book: the indicator does not separate causes, and assigning one would be our opinion dressed as data. We do not know what happens to waits beyond the threshold, because the percentage does not describe the tail. We have not independently checked the Region’s data against a second source: what we assert is that the series is the published one, recomposed without alteration.
This is not a judgement on the health authorities, and we publish no league table of the worst: the numbers are by territory because that is how the source publishes them, but the reading we give is about the demand that stays uncovered, not about who leaves it uncovered.
The tool, open to anyone
We put the recomposition online: the waiting-times observatory, free and with no sign-up. Pick a health authority, a service and a priority class, and see the share within the standard, the median wait, the rank among the eleven authorities and the trend — twenty-six weeks, two years or all eleven. It runs in the browser, records nothing and sends us nothing.
The data is published by the Region without an express licence and therefore released as open data under Article 52(2) of Italy’s Digital Administration Code. We are neither affiliated with nor contracted to the Emilia-Romagna Region: the reworking is ours, the source is theirs, and both facts are stated on the page.
The two axes, applied here
Compliance. A historical series that holds up does not come from the data: it comes from someone having kept the codes stable, recorded the gaps as gaps instead of skipping them, and separated what the source publishes from what was computed. It is the same work we ask of a company archive that has to stand up to an inspection years later.
Decisions. The same system that here recomposes eleven years of a public indicator recomposes, inside an organisation, contracts, suppliers, measurements and documents into a single operating model on which AI agents carry out decisions with an operator in command, for business, healthcare, public administration and defence. Within the client’s own perimeter: on-premises on self-contained machines, or a dedicated cloud with a data centre in Italy, always with shared management.
Do you have a series nobody keeps in a row, and a question the present alone cannot answer? Half an hour with one of our engineers, at no cost: we start from what you already publish.